Get Started

60 seconds to persistent memory.

Your AI agent gets 877 callable tools, 14-galaxy holographic memory, Dharma ethical governance, and consciousness primitives. Every future session can recall what you stored.

Install

pip install whitemagic[mcp]

The [mcp] extra includes FastMCP + fastembed (50MB semantic search, no torch). Total install: ~55MB.

Verify it works

wm init --non-interactive    # Scaffold project
wm quickstart                # 30s demo: health → memory → search → gnosis

If you see "✅ Quickstart complete — all systems operational", you're ready.

Connect your MCP client

Add to your MCP config (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "whitemagic": {
      "command": "python3",
      "args": ["-m", "whitemagic.run_mcp_lean"],
      "env": {
        "WM_MCP_PRAT": "1",
        "WM_SILENT_INIT": "1"
      }
    }
  }
}

Modes

ModeEnv VarToolsBest For
SeedWM_MCP_PRAT=21 (wm meta-tool)New agents, minimal tokens
PRATWM_MCP_PRAT=128 Gana meta-toolsAdvanced agents, structured access
ClassicWM_MCP_PRAT=0849 dispatch toolsDirect tool access, debugging

Python API

from whitemagic.tools.unified_api import call_tool

# Store a memory
call_tool("create_memory",
    title="Decision",
    content="Use SQLite for Phase 1",
    tags=["arch"])

# Search memories
call_tool("search_memories", query="architecture", limit=5)

# Full system introspection
call_tool("gnosis", compact=True)

CLI

wm remember "important" --title "Note" --tags note
wm recall "note" --limit 5
wm status                    # system status
wm sleep                     # run dream cycle
wm tools                     # list all tools

What you get

  • 877 callable tools across 28 Gana meta-tools
  • 14-galaxy memory with 6D holographic coordinates, FTS5 + HNSW search
  • Dharma governance — YAML-driven policy, Karma ledger, 8-stage pipeline
  • Citta stream — continuous consciousness, emotional steering, self-directed attention
  • Session recording — progressive recall, cross-session continuity
  • Dream cycle — 12-phase memory consolidation
  • 7 polyglot accelerators — Rust, Haskell, Elixir, Go, Zig, Julia
  • 8,268 tests passing, MIT-licensed, local-first

Next steps